Switch Testing

For testing any changes for the xrootd managers:

  1. Roll out the change only on manager02

  2. systemctl stop xrootd@{unified,tpc} cmsd@unified keepalived on manager01

  3. leave the system running for a day

  4. in case any failures are observed:
    systemctl start xrootd@{unified,tpc} cmsd@unified keepalived

  5. otherwise roll out on manager01 as well, then restart the above services on manager01

Rationale

The primary manager is the only one doing most of the work. (manager01 in this case). stopping manager01 switches manager02 into being primary, and keepalived shifts the floating IPs onto it as well. In case any errors occur, reversion to a known working state can happen within seconds